August 1, 2018.
What if there was a way to detect undeclared (Clandestine) facilities using data sets from a nuclear facility? The Statistical Modeling and Experimental Design group at the Pacific Northwest National Laboratory are developing the Modeling and Inference for Remote Sensing (MIRS) model. The model uses two deterrence scenarios within the nuclear fuel cycle to detect undeclared facilities. The two scenarios are used in an agent based nuclear fuel cycle simulator to produce the declared and undeclared feed, tails assay, and sink inventory in kilograms of Uranium. In the first scenario, natural uranium is being diverted from the conversion plant to an undeclared facility. In the second scenario, low enriched uranium is diverted from the declared enrichment facility to an undeclared facility. By changing the value in each variable in the enrichment archetype, the graphs will display the behavior of diverted uranium. The values were changed, in the code, one variable at a time for each scenario in both the declared and undeclared enrichment facilities. By using a set of ranges provided by an in-house expertise, a parametric search was conducted. From the simulations ran, detection of diverted uranium can be seen easily in the declared sink inventory graph of scenario 2. However, there are cases in scenario 1 where detection of uranium is hard to detect. Deeper parametric search needs to be conducted in order to assess detection in the declared sink inventory in scenario 1. Once completed, machine learning time series classification algorithms will be implemented to classify declared and undeclared enrichment facilities.
Romarie Morales Rosado
Pacific Northwest National Laboratory (PNNL)
The 2018 STEM Teacher and Researcher Program and this project have been made possible through support from Chevron (www.chevron.com), the National Marine Sanctuary Foundation (www.marinesanctuary.org), the National Science Foundation through the Robert Noyce Program under Grant #1836335 and 1340110, the California State University Office of the Chancellor, and California Polytechnic State University in partnership with Pacific Northwest National Laboratory. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the funders.